Association rule mining (ARM) is a fundamental technique in knowledge discovery and intelligent decision analysis, yet traditional algorithms such as Apriori and FP-Growth suffer from exponential candidate growth, high memory usage, and poor adaptability in dynamic environments. Metaheuristic and swarm intelligence algorithms have recently emerged as promising alternatives, but most existing methods still face two limitations: insufficient balance between global exploration and local exploitation, and low robustness when data are noisy or evolving. To address these challenges, we propose a multi-strategy enhanced hybrid Salp Swarm Algorithm (Hybrid SSA) for association rule mining. The proposed framework integrates four complementary strategies: Latin hypercube initialization to improve population diversity, Lévy flight perturbations to strengthen global search, adaptive parameter control to dynamically adjust exploration and exploitation, and elite retention to preserve high-quality rules. This combination enables Hybrid SSA to maintain scalability, robustness, and interpretability under large-scale and time-varying conditions. We provide theoretical analysis establishing convergence and diversity guarantees, and validate the method through comprehensive experiments on benchmark datasets (Retail, Mushroom, and Telecom). Results demonstrate that Hybrid SSA significantly outperforms Apriori, FP-Growth, and genetic algorithm baselines in terms of rule quality (lift and confidence), convergence speed, and robustness under noisy environments. The contributions of this work are threefold: (1) designing a hybrid SSA framework unifying multiple enhancement strategies; (2) providing theoretical and empirical evidence of improved convergence and resilience; and (3) demonstrating superior performance in real-world decision analysis tasks, confirming the potential of Hybrid SSA for scalable and adaptive association rule mining.
Lin et al. (Wed,) studied this question.